The race to own the AI backbone of banking is heating up, with Anthropic's recent launch of financial services-focused AI agents being a pivotal moment. This move signifies a shift in the AI vendor landscape, where companies are no longer just selling into banking but are now embedding themselves within the industry. Anthropic's AI agents, designed for tasks like pitchbook preparation and underwriting support, are already being adopted by major institutions like Goldman Sachs, Visa, Citi, and AIG. This trend is not isolated; OpenAI is also partnering with PwC to bring agentic AI to finance, aiming to reimagine the CFO's office. The competition is intensifying, with AI firms vying to become integrated layers within banking infrastructure, compliance operations, fraud systems, and treasury management. This operational positioning is deeper than retail chatbot adoption, as it touches upon systems that manage risk, capital allocation, and regulatory obligations. The challenge for financial institutions is not just technical integration but also ensuring that external AI systems can operate within environments governed by audit requirements, cybersecurity controls, model-risk standards, and supervisory review. Regulators are taking notice, with the Federal Reserve Vice Chair for Supervision, Michelle Bowman, warning that AI capabilities are advancing quickly enough to require updated supervisory approaches. The issue of operational dependence and risk is also emerging, as early deployments of AI systems focused on customer service and employee productivity are now being extended to automate internal review layers, including transaction monitoring, sanctions screening, and commercial loan documentation. While these initiatives may improve operational efficiency, they also deepen institutional dependence on a relatively narrow group of AI and cloud providers. The central question for financial institutions is no longer whether AI can improve productivity but rather how to embed these systems inside regulated financial environments without compromising cybersecurity, operational continuity, and compliance standards. This is a complex challenge, and the race to own the AI backbone of banking is far from over.
